The roles of the cerebellum and basal ganglia in timing and error prediction

The roles of the cerebellum and basal ganglia in timingand error prediction

Jean-Claude Dreher and Jordan GrafmanCognitive Neuroscience Section, National Institute of Neurological Disorder and Stroke, Building 10, Room 5C205,

MSC 1440 Bethesda, Maryland 20892±1440, USA

Keywords: attention, fMRI, predictability, striatum, switching

Abstract

Recent evidence that the cerebellum and the basal ganglia are activated during the performance of cognitive and attention tasks

challenges the prevailing view of their primary function in motor control. The speci®c roles of the basal ganglia and thecerebellum in cognition, however, have been dif®cult to identify. At least three functional hypotheses regarding their roles have

been proposed. The ®rst hypothesis suggests that their main function is to switch attentional set. The second hypothesis states

that they provide error signals regarding stimuli or rewards. The third hypothesis is that they operate as an internal timing system,providing a precise representation of temporal information. Using functional magnetic resonance imaging, we tested these three

hypotheses using a task-switching experiment with a 2 3 2 factorial design varying timing (random relative to ®xed) and task

order (unpredictable relative to predictable). This design allowed us to test whether switching between tasks, timing irregularity

and/or task order unpredictability activate the basal ganglia and/or the cerebellum. We show that the cerebellum is primarilyactivated with timing irregularity while the anterior striatum is activated with task order unpredictability, supporting their distinctive

roles in two forms of readjustment. Task order unpredictability alone, independent of reward delivery, is suf®cient to induce

striatal activation. In addition, activation of the cerebellum and basal ganglia were not speci®c to switching attention becausethese regions were both activated during switching between tasks and during the simultaneous maintenance of two tasks without

switching between them.

Introduction

The basal ganglia and the cerebellum are known to be involved in

motor control because of the motor de®cits resulting from their

dysfunction. This motor function has been supplemented in recent

years by numerous studies showing that the basal ganglia and the

cerebellum are also involved in a wide range of nonmotor, cognitive

tasks (Leiner et al., 1991; Ivry, 1996; Middleton & Strick, 2000).

Although several studies have contrasted the roles of the cerebellum

and of the basal ganglia in motor control and learning (Jueptner &

Weiller, 1998; Doya, 2000), it has been dif®cult to distinguish their

speci®c roles in cognitive performance. Three main hypotheses have

been offered.

The ®rst hypothesis is that the cerebellum and basal ganglia

mediate switching attentional set (Akshoomoff & Courchesne, 1992;

Owen et al., 1993; Hayes et al., 1998; Le et al., 1998). Indeed, both

cerebellar lesions patients (Courchesne & Allen, 1997) and

parkinsonian patients showed impaired performance in switching

attention set or switching between tasks (Gotham et al., 1988;

Eslinger & Grattan, 1993; Owen et al., 1993; Hayes et al., 1998;

Cools et al., 2001b). FMRI studies also support the view that the

lateral cerebellar hemisphere is activated in shifting attention (Le

et al., 1998), but no speci®c basal ganglia activation was found for

switch as compared to repeat trials during task switching (Dove et al.,

2000; Sohn et al., 2000).

The second hypothesis is that the cerebellum and the basal ganglia

code prediction errors (Fiez et al., 1992; Kawato & Gomi, 1992;

Schultz & Dickinson, 2000; Ivry & Fiez, 2001). Evidence that the

cerebellum codes prediction errors was ®rst proposed for the motor

domain, based on the fact that the cerebellum modi®es its output with

unexpected sensory input (Marr, 1969; Albus, 1971; Kawato & Gomi,

1992). This error-correction function has been extended to the

cognitive domain, the cerebellum providing a more general error-

correction role, in modifying internal thought (Bracke-Tolkmitt et al.,

1989; Canavan et al., 1994; Courchesne & Allen, 1997; Hikosaka

et al., 1999; Tamada et al., 1999; Doya, 2000; Imamizu et al., 2000;

Blakemore et al., 2001; Ivry & Fiez, 2001). Evidence that the basal

ganglia code an error prediction is mainly based on the fact that

monkey's striatum codes a prediction error of reward delivery

(Schultz, 2000; Schultz & Dickinson, 2000). A recent fMRI study in

humans con®rms this hypothesis, the ventral striatum being activated

by reward unpredictability (Berns et al., 2001). However, it is unclear

whether striatal activity necessarily requires reward delivery or

whether the crucial factor is unpredictability itself.

The third hypothesis is that the basal ganglia and the cerebellum

operate as an internal timing system, providing the precise representation

of temporal representation across various tasks. Cerebellar patients have

dif®culty with perceptual tasks requiring precise timing (Ivry, 1996,

1997) and Parkinson's disease patients often underestimate time

intervals (Pastor et al., 1992; Harrington et al., 1998). Direct evidence

that the cerebellar neocortex is essential for timing the interval between

stimulus and response in eyeblink conditioning has been provided by

extensive neurophysiological and lesion studies in the rabbit.

Correspondence: Jordan Grafman, PhD, as above.E-mail: [email protected]

Received 3 May 2002, revised 28 July 2002, accepted 6 August 2002

doi:10.1046/j.1460-9568.2002.02212.x

European Journal of Neuroscience, Vol. 16, pp. 1609±1619, 2002 ã Federation of European Neuroscience Societies

Although cerebellar function is also associated with associative

learning and sensorimotor coordination (Marr, 1969; Albus, 1971;

Thach, 1998; Miall et al., 2001), we concentrate on testing the three

theories mentioned above because they concern both the basal

ganglia and the cerebellum. The goal of this study was thus to test

whether activation of the cerebellum and basal ganglia are speci®c to

switching attention between tasks, to error prediction or to timing

irregularity. To address these questions, we conducted a task-

switching fMRI experiment using a 2 3 2 factorial design varying

timing (®xed relative to random) and task order (predictable relative

to unpredictable).

Materials and methods

Subjects

Eight healthy subjects (mean age 25 years; range 20±31) with at least

a high school education were recruited following procedures

approved by the Institutional Review Board. All subjects were native

speakers of English and strongly right-handed, as measured by the

Edinburgh handedness inventory (mean score 90.5). None of the

subjects presented with an abnormal neurological history. One or two

days before the MR session, subjects participated in a behavioural

testing session during which they were trained to perform each of the

tasks and were required to attain an overall accuracy score > 90% to

participate in the fMRI experiment.

Stimuli and task conditions

Subjects responded to visually presented letters (vowels or conson-

ants, either in lower or upper case, in red or in green) by pressing

response buttons with their right or left hand (Fig. 1). There were

eight conditions, each cued by a distinct visual instruction, consisting

of two conditions used for baseline (Task A, vowel-consonant

discrimination; Task B, case discrimination), four tasks switching

conditions (obtained by crossing the task order and timing factors)

and two conditions used for control (Union task, A or B) (see full

description below). The tasks were administered in six scanning runs.

Each condition was included once in each run as a block of 24 trials

for a total of 192 trials/run. Each run was pseudo-randomly ordered

so that each condition appeared at different serial positions within a

run and two conditions never appeared twice in immediate succession

to prevent confounding order effects. Control tasks or tasks used for

baseline alternated with switch conditions.

Sustained attention tasks (Baseline)

In the `vowel±consonant' condition, subjects had to press the right

button if the letter was a vowel and the left if the letter was a

consonant. In the `case discrimination' condition, subjects had to

press the right button if the letter was in upper case and the left if the

letter was in lower case. In both of these conditions, the colour of the

letters was irrelevant. The baseline was composed of the mean of

these two discrimination tasks [(A + B)/2]. In both of these two

conditions, the colour of the stimuli (appearing every 2.5 s) alternated

every three letters.

Switching tasks

In the `switching' conditions, subjects had to take into account the

colour of the letters to know which task to perform. If the letter was

red, subjects had to perform the vowel±consonant condition (i.e. to

respond with the right button if the letter was a vowel and with the

left if it was a consonant). Conversely, if the letter was in green,

subjects had to perform the case discrimination task (i.e. to respond

with the right button if the letter was in upper case and with the left if

it was in lower case). There were four different switch conditions:

`®xed predictable', `random predictable', `®xed unpredictable',

`random unpredictable'). In the ®xed conditions the stimuli appeared

every 2.5 s whereas, in the random condition, the timing between two

stimuli was pseudo-randomized (2.5 s 6260, 6390, 6510 ms). In

the predictable condition, a switch between tasks occurred every three

stimuli and was pseudo-randomized in the unpredictable condition.

Across each condition, the number of red/green letters (12 of each),

of switches (11) and of left/right responses (12 of each) was

consistently maintained.

Control condition

The mean of the following two conditions was used to control for the

active maintenance of the two task sets without switching between

them. In both of these conditions, subjects had to press the right

button if the letter was a vowel or was in upper case (and if both were

true) and the left button otherwise (Union condition). In the ®rst

control condition stimuli appeared every 2.5 s while, in the second

control, the timing between two stimuli was pseudo-random every

2.5 s 6 260, 390 or 510 ms. In both of these control conditions, the

colour of the letter was irrelevant and changed pseudo-randomly.

Stimulus duration was 500 ms in all conditions.

This design allowed us to test the three hypotheses mentioned in

the introduction. If the cerebellum and/or the basal ganglia are

speci®c to switching attention between tasks, they should be more

activated by task switching than by maintaining two task sets without

switching between them. Furthermore, if the cerebellum and/or the

basal ganglia are speci®c to timing adjustment, they should be

activated by the main effect of timing irregularity (random > ®xed

timing conditions) because the uncertainty of the temporal relation-

ship between events is maximal in the random condition.

Behaviourally, a timing operation may be revealed either by slower

reaction times (RTs) in the random compared to the ®xed timing

conditions or by a reduction in RTs with an increasing time lag in the

random timing condition. The latter possibility is likely to occur if

subjects use the probabilistic information conveyed by the passage of

time to predict the likelihood of stimulus presentation (i.e. the more

time the subjects have, the more probable the stimulus occurrence). In

both cases, the main effect of timing irregularity (random > ®xed

timing) should reveal brain regions that attempt but fail to predict the

timing of stimulus presentation [at least for short interstimulus

intervals (ISIs)]. Conversely, the main effect of task order unpredict-

ability (unpredictable > predictable task order) should reveal brain

regions activated in error prediction, because more adjustment to the

current task is needed when task order is unpredictable than when it is

fully predictable. Indeed, subjects may constantly try to anticipate the

following task. Thus, when task order is unpredictable, discordance

between subject's expectation and the effective task to perform will

result in error correction. In contrast, when task order is predictable,

routine preparation for the whole sequence of tasks is possible and no

error signal needs to be emitted.

Apparatus

Stimuli were generated by a PC computer using the Expe6 software

package (ftp://ftp.lscp.ehess.fr/pub/expe6/) and projected onto a

screen at the subjects' feet. Subjects viewed the stimuli through a

mirror attached to the head coil. Subjects' behavioural responses were

recorded when they pressed one of two MRI-compatible response

buttons held in each hand.

1610 J.-C. Dreher and J. Grafman

ã 2002 Federation of European Neuroscience Societies, European Journal of Neuroscience, 16, 1609±1619

Data analysis

Behavioural analysis

RT and percentage of errors were analysed with a MANOVA including

the two factors of Task Order (predictable vs. unpredictable) and

Timing (®xed vs. random).

Imaging data

High-resolution structural images were obtained using a standard 1.5

GE whole-body signa scanner with an RF head coil. For each subject,

six time series of 180 whole-brain images (four ®rst images removed)

were obtained with a gradient-echo, echo-planar scanning sequence

(TR 3 s, TE 40 ms, ¯ip angle 90°; FOV 24 cm, acquisition matrix

64 3 64, 22 axial slices, thickness 6 mm). Each run was pseudo-

randomly ordered according to a latin square design so that each

condition appeared only once at different serial positions within a run

and that baseline/control and switch conditions alternated. The order

of runs was also counterbalanced across subjects. Using SPM96 with

modi®ed memory-mapping procedures, for each subject, the series of

functional images for the six runs was realigned, normalized to the

Montreal Neurological Institute template, smoothed with a Gaussian

®lter (10-mm FWHM kernel), and ®nally scaled across scans. Then,

the data from all subjects were pooled and statistical parametric maps

were computed from local MR signals using linear multiple

regression with conditions, modelled as two temporal basis functions,

and runs as covariates (Friston et al., 1991). Only regions formed by

more than four adjacent active voxels (voxel size 4 mm3) were

reported (extend threshold P < 0.5). In accordance with our hypo-

theses, only cerebellar and subcortical regions are reported (Z > 3.09,

i.e. P < 0.001, uncorrected for multiple comparisons).

Mean values are quoted 6 SEM.

Results

Behavioural data

Behavioural results are depicted in Fig. 2. There was a signi®cant

response time cost (RT difference between switch and no switch

trials) in all task-switching conditions [F1,7 = 32.2; P < 0.001] (main

effect of Switch). Furthermore, performance was signi®cantly worse

when the order of the tasks was unpredictable than when it was

predictable (RT = 932.2 6 130.1 vs. 818 6 142.9 ms, F1,7 = 14.4,

P < 0.01; 4.7 6 3.1% of errors vs. 3.9 6 3.4%, F1,7 = 5.7, P < 0.05;

main effect of Task Order). Finally, no main effect of Timing was

found: ®xed timing, compared to random timing, did not signi®cantly

reduced the error rate (3.8 6 3.4% vs. 4.8 6 3.3%, F1,7 = 2.0,

P = 0.19) nor the RTs (865.6 6 138.1 vs. 884.6 6 123.8 ms,

F1,7 = 2.2, P = 0.18). No interaction reached statistical signi®cance.

The mean RTs of the individual tasks used for baseline and control

were: vowel/consonant, 717.3 6 100.1 ms; case discrimination,

656.5 6 89.8 ms; ®rst union, 778 6 115.2 ms; second union,

747.1 6 79.2 ms.

It is possible that the different time lags between stimuli in the

random timing conditions induce different types of behaviour (e.g.

subjects may be better prepared with long than with short ISIs) which

may not be observed by directly comparing the ®xed to the random

timing conditions. In order to better understand the contribution of the

various time lags in random timing, we thus performed an additional

three-factors repeated-measures ANOVA only for the two random-

timing switching conditions, which included task order (Predictable

vs. Unpredictable), time lag (®ve intervals of 408 ms from 1480 to

3520 ms) and switch (no switch vs. switch trials). As before, we

found a main effect of Switch (F1,7 = 15.3, P < 0.01) and a main

effect of Task Order (F1,7 = 9.5, P < 0.05). There was a trend toward

a reduction of RTs with time lag (F4,28 = 2.2, P = 0.09); this trend

became signi®cant (F4,56 = 3.2, P < 0.05) in another analysis adding

seven pilot subjects. No interactions were found between these

factors. Error rates did not show any signi®cant main effect or

interactions.

To ensure that the control condition effectively induced the

subjects to maintain two tasks in memory without switching between

them, we directly compared the control and the switching conditions.

RTs were signi®cantly slower in the task-switching conditions

averaged together (mean 875.0 6 129.9 ms) than in the control

condition (mean 763.2 6 96.1 ms; F1,7 = 34.8; P < 0.001), indicat-

ing that the switching conditions demanded additional cognitive

processes over those needed for the control task. Furthermore,

subjects did maintain two task-sets simultaneously in memory in the

control condition because, if subjects coded the control condition as a

single stimulus±response mapping, they should have directly associ-

ated the lower case consonants to a left motor response and other

letters to a right button press. However, RTs were not faster for lower

case consonants than for other letters (vowels and upper case

consonants) (F1,7 = 0.95; P = 0.4).

Imaging data

Tables 1 and 2 present results from all the brain regions activated in

the different conditions. In the following, we focus on subcortical and

cerebellar activation to test our hypotheses. First, we investigated

which cerebellar and subcortical regions were signi®cantly activated

in all switching conditions averaged together relative to baseline

(Z > 3.09, i.e. P < 0.001, uncorrected). Activation was found in the

cerebellar hemispheres (lobules VI, VIIIA, VIIIB, Crus I, in

Schmahmann's nomenclature; Schmahmann et al., 1999), the vermis

(lobules VII, VIII) (Fig. 3), the thalamus and the caudate nucleus

(Fig. 4). We plotted the activation of each task-switching condition

separately relative to baseline at the peak of activation of each of

these brain regions (Z > 3.09, i.e. P < 0.001, uncorrected) (Figs 3

and 4, bottom).

In order to control for the higher working memory load necessary

to keep both task sets active during the task-switching conditions, we

also examined the voxels commonly activated by each task-switching

condition relative to the control condition, which required subjects to

maintain two task-sets in memory without switching between them.

This was done by selecting the voxels activated in all task-switching

conditions averaged together relative to baseline (Z > 3.09,

P < 0.001, uncorrected for multiple comparisons) which were also

activated in each individual task-switching condition compared

separately to baseline (Z > 2.33, P < 0.01, uncorrected). None of

the brain regions mentioned above (caudate nucleus, thalamus and

cerebellum) were signi®cantly activated in this comparison (Z < 1.7,

P > 0.05, uncorrected). Furthermore, all these brain regions were

signi®cantly activated in the control condition relative to baseline

(Z > 3.09, i.e. P < 0.001, uncorrected), indicating that they were not

speci®cally activated during task switching (graph bars in Figs 3

and 4).

The main effect of task-order unpredictability (subtraction of the

two task-switching conditions with predictable order from the two

with unpredictable task order) activated the anterior putamen

bilaterally (Z > 3.09, i.e. P < 0.001, uncorrected) (Table 1, Fig. 5).

No signi®cant activation was found in the cerebellum.

The main effect of timing irregularity (subtraction of the two task-

switching conditions with ®xed timing from those with random

timing) was to activate the right posterior cerebellar hemispheres (VI/

Cerebellar and striatal functions in timing and error prediction 1611


VIIB/Crus I) (Schmahmann et al., 1999) and the dentate nucleus

(Z > 3.09, i.e. P < 0.001, uncorrected) (Table 1, Fig. 6). No signi®-

cant activation was found in the basal ganglia.

Brain regions showing interactions between the timing and task-

order factors (Z > 3.09, i.e. P < 0.001, uncorrected) activated the

right head of the caudate nucleus and the posterior putamen

bilaterally (Table 1, Fig. 7). No interactions were found in the

cerebellum.

Finally, we investigated whether the combination of task order and

timing unpredictability activated speci®c subcortical regions relative

to baseline (Z > 3.09, i.e. P < 0.001, uncorrected). We found that the

task-switching condition with unpredictable task order and random

timing speci®cally activated the substantia nigra (Fig. 8).

Functional connectivity analysis

In order to better understand what the basal ganglia/cerebellum do

with task order/timing information, we performed a functional

connectivity analysis on our data. This analysis allows us to point

out which brain regions correlate with the cerebellum and the basal

ganglia in each task-switching condition. For each subject, we

measured the mean blood oxygenation level-dependent (BOLD)

TABLE 1. Foci of activations in the different statistical contrasts

All task-switching conditionsaveraged together vs. baseline

Main effect ofirregular timing

Main effect ofunpredictable task order

Anatomical structure (Brodmann's area) x y z Z-value x y z Z-value x y z Z-value

CerebellumLeft cerebellar hemisphere ±28 ±68 ±40 6.72

±44 ±64 ±28 6.68Right cerebellar hemisphere 32 ±64 ±32 3.71 28

20±60±40

±32±44

4.965.56

Vermis (VIIB/VIIIB) 0 ±68 ±32 5.80Basal ganglia

Left Caudate nucleus ±16 ±16 24 5.15Right Caudate nucleus 16 4 20 5.13Left Putamen ±20 4 0 3.66Right Putamen 20

1612

168

±4

0±4±8

4.604.513.83

ThalamusLeft thalamus (vl) ±12 ±16 8 4.78Right thalamus (dm) 28 ±12 8 4.79

FrontalL sFG (BA 6) ±28 4 56 7.56L mFG (BA 8/9/44) ±40 20 32 7.18 ±44 16 32 5.79R mFG (BA 8/9/46) 48 16 40 7.77

60 28 32 6.47R iFG (BA 44/45) 36 24 20 6.21L Somato-motor area (BA 4) ±28 ±16 60 5.16R Somato-motor area (BA 4) 20 ±16 68 6.21Parietal

L IPS (BA 7/40) ±36 ±56 52 8.36 ±32 ±56 48 6.10R IPS (BA 7/40) 36 ±60 44 8.05 40 ±48 48 5.74

Temporal gyrusL iTG (BA 20) ±52 ±28 ±12 6.64R iTG (BA 37) 52 ±56 ±4 7.54 52 ±56 ±4 7.54

InsulaL INS ±20 24 ±4 5.59R INS 28 28 4 5.37L m occip. gyrus (BA 18/19) ±28 ±88 4 6.19R m occip. gyrus (BA 18/19) 28 ±84 ±8 5.38

All areas were signi®cant at P < 0.001 (uncorrected for multiple comparisons); x, y, and z are standardized stereotaxic coordinates of Talairach and Tournoux.Abbreviations: L, left; R, right. sFG, superior frontal gyrus; mFG, middle frontal gyrus; IPS, intra-parietal sulcus; INS, insula; iTG, inferior temporal gyrus; sTG,superior temporal gyrus; Somatosens. cx, somatosensory cortex; occip. gyrus, occipital gyrus; vl, ventro-lateral, dm, dorso-medial.

TABLE 2. Areas revealing interactions between task order and timing

Talairach coordinates

Anatomical structures x y z Z-value

Frontal cortexRight PM/Motor cx (BA 4/6) 52 ±8 28 6.5Left Motor cx (BA 4) ±44 ±12 32 6.1Right mFG (BA 9/46) 40 28 28 6.5Left mFG (BA 9/46) ±36 28 24 5.8Left fronto-polar cx (BA 10) ±28 48 8 5.2Pre-SMA ±4 12 56 4.9

Cingulate cortexAnterior cingulate (BA 24) 4 28 16 5.4

Basal gangliaRight Caudate nucleus 16 12 4 3.79Left Putamen ±24 ±8 ±4 4.08Right Putamen 20 ±4 4 3.68

Temporal gyrusRight mTG (BA 19/37) 40 ±72 4 5.7Left mTG (BA 37) ±52 ±36 ±8 4.8

BA, Brodmann's area; mFG, medial frontal gyrus; mTG, middle temporalgyrus; PM, pre-motor cortex; x, y, z, Talairach coordinates. All areas weresigni®cant at P < 0.001 (uncorrected for multiple comparisons).



FIG. 1. (a) Table describing the 2 3 2 factorial design. The two factors ofinterest were timing (Fixed, 2.5 s; Random, 2.5 s 6 260, 390 or 510 ms)and task order (Predictable, switch on every second trial; Unpredictable).The factors were crossed to produce four distinct task-switching conditions:Fixed timing Predictable order (FP), Fixed timing Unpredictable order (FU),Random timing Predictable order (RP), Random timing Unpredictable order(RU). (b) Stimuli and timing. Subjects responded to visually presentedletters by pressing response buttons with their right or left hand. Subjectshad to switch between two letter-discrimination tasks depending upon thecolour of the letter. If the letter was red, subjects performed a vowel±consonant discrimination task (Vowel, right; Consonant, left). If the letterwas green, subjects performed a case-discrimination task (Upper case, right,Lower case, left). Two (FP and RU) of the four possible tasks switchingconditions are represented in examples. (Left) Stimuli in the predictabletask order condition with ®xed timing. (Right) Stimuli in the unpredictabletask order condition with random timing.

FIG. 2. (a) Mean response time (left) and error rates (right) averaged acrosssubjects in the various tasks switching conditions for the repeat and switchtrials. Mean response time was examined for correct response trials only.(b) Response time for each of the ®ve intervals grouping the different ISIsof the two task-switching conditions with random timing. Errors barsrepresent SEM. The random timing condition 2.5 s 6 260, 390 or 510 mscreates the following ISIs: 1480, 1600, 1720, 1730, 1850, 1980, 2250, 2370,2380, 2500, 2620, 2630, 2750, 3020, 3150, 3270, 3280, 3400 and 3520 ms.

FIG. 3. (a) Data for cerebellar regions signi®cantly activated in all tasksswitching conditions relative to baseline (mean of the two discriminationtasks) were superimposed on the mean structural MRI slices averagedacross subjects. Slices are shown in neurological convention. The Z-valuesfor the regional maxima are listed in Table 1. (b) Percentage of signalchange relative to baseline for each task-switching condition and for thecontrol condition at the peak of activation in the cerebellar hemispheres.Error bars indicate SEM.



signal in each task-switching condition relative to baseline at the peak

of activation of the following brain regions activated by all task-

switching conditions relative to baseline (Table 1): the dorsolateral

prefrontal cortex (DLPFC) (x,y,z = 48,16,40; x,y,z = ±28,4,56), the

intraparietal sulcus (IPS) region (x,y,z = ±36,56,52; x,y,z = 36,60,44),

the pre-supplementary motor area (pre-SMA) (x,y,z = ±4,16,52), the

caudate nucleus (x,y,z = ±16,16,24; x,y,z = 16,4,20) and the cerebel-

lum (x,y,z = ±28,68,40; x,y,z = 32,64,32; x,y,z = 0,68,32). In each

task-switching condition, we then performed correlations between

each of these brain regions. We focused on brain regions correlating

with the caudate nucleus in the unpredictable task order conditions

and with brain regions correlating with the cerebellum in the random

timing conditions (Fig. 9). Figure 9a represents the pattern of

activation in the different task-switching conditions relative to

baseline (Z > 3.09, i.e. P < 0.001, uncorrected) and Fig. 9b (top)

represents the matrix of interregional correlations (R2). In order to

clarify the results, we also collapsed together the data from left and

right hemispheres in these brain networks (Fig. 9b, bottom).

When task order was unpredictable (and timing was ®xed or

random), the caudate activation positively correlated with activation

of the DLPFC and the cerebellum. In addition, when both task order

was unpredictable and timing was random, the caudate activation

FIG. 4. (Top) Data for striatal and thalamic regions signi®cantly activated inall tasks switching conditions relative to baseline were superimposed on themean structural MRI slices averaged across subjects. Slices are shown inneurological convention. The Z-values for the regional maxima are listed inTable 1. (Bottom) Percentage of signal change relative to baseline for eachtask-switching condition and for the control condition in the caudatenucleus and thalamus. Error bars indicate SEM.

FIG. 5. (Left) Main effect of task order unpredictability superimposed on anormalized coronal MRI slice of one subjects. (Right) Percentage of signalchange relative to baseline for each task-switching condition in the rightanterior putamen. Error bars indicate SEM.

FIG. 6. (Top) Main effect of timing irregularity superimposed on the meanstructural MRI slices averaged across subjects. The Z-values for the regionalmaxima are listed in Table 1. (Bottom) Percentage of signal change relativeto baseline for each task-switching condition in the right cerebellarhemisphere. Error bars indicate SEM.

FIG. 7. Interactions between task order and timing superimposed onnormalized structural MRI slices averaged across subjects. Percentage ofsignal change relative to baseline for each task-switching condition isshown at the peaks of striatal activation. Error bars indicate SEM.



correlated with the pre-SMA and the IPS. Thus, the caudate nucleus

communicates information relative to task order unpredictability to a

number of cortical regions.

When timing was random (and task order was predictable or

unpredictable), the cerebellum activation positively correlated with

the DLPFC, the IPS and the caudate activation. In addition, when

both timing and task order were unpredictable, the cerebellum

activation positively correlated with the pre-SMA activation.

Discussion

Our results show that both the cerebellum and the striatum were

activated bilaterally by all tasks switching conditions relative to

baseline (Figs 3 and 4). These activations cannot be attributed to

motor function because identical motor responses were required for

the task-switching and baseline conditions. Activations of the

cerebellum and basal ganglia do not support the switching attention

hypothesis because these brain regions were also activated by a

control condition requiring maintenance of two task-sets without

switching between them (Figs 3 and 4). Rather, our results show that

the cerebellum and the basal ganglia are differently sensitive to the

unpredictability of task order and timing during the processing of

sequences of tasks. The anterior striatum was activated by the main

effect of task order unpredictability (Fig. 5). In contrast, the right

posterior cerebellar hemisphere and dentate nucleus were activated

by the main effect of timing irregularity (Fig. 6).

Brain regions activated with task order unpredictability or timing

irregularity may, in part, be due to a form of oddball response.

However, it should be kept in mind that in classical oddball

paradigms both the timing and the type of stimulus (target or

distractor) are unpredictable (and thus confounded) while our study

dissociated the two components of timing and task order unpredict-

ability. Event-related potential (ERP) studies of the oddball paradigm

have revealed that rare target stimuli generate the P3b event-related

potential (Clark et al., 2000). Moreover, oddball paradigms that have

investigated the neural responses to infrequent stimuli in the auditory

or visual domains found a large degree of spatial overlap primarily in

the bilateral cerebellum and the frontal and parietal areas, as well as

in the supramarginal gyrus, frontal operculum and insular cortex

bilaterally (Linden et al., 1999; Clark et al., 2000; Stevens et al.,

2000). Although we limit our discussion to the cerebellum and the

basal ganglia in the present paper, we have previously described and

discussed the roles of distinct prefrontal regions and of the

intraparietal cortex with task order predictability/unpredictability

and timing regularity (Dreher et al., 2002).

Our results that the cerebellum and basal ganglia are not speci®c

for switching attention are in accordance with the fact that cerebellar

patients are not impaired on attention tasks that require rapid visual

orienting between spatial positions (Dimitrov et al., 1996; Helmuth

et al., 1997; Yamaguchi et al., 1998). Although early studies reported

attentional de®cits in cerebellar patients (Courchesne et al., 1994;

Courchesne & Allen, 1997), it has recently been proposed that these

de®cits might be secondary to the coordination of motor responses

(Ravizza & Ivry, 2001; Bischoff-Grethe et al., 2002). The observation

that the basal ganglia are not speci®c for switching between tasks is

also provided by several event-related fMRI studies directly

comparing switch to repeat trials (Kimberg et al., 2000; Sohn et al.,

2000). Furthermore, early-stage medicated Parkinson's disease

patients performed normally in switching attention between tasks in

which verbal responses were given (Rogers et al., 1998). Although

patients with Parkinson's disease have also been reported to have

impaired performance in switching attention set (Taylor et al., 1986;

Gotham et al., 1988; Eslinger & Grattan, 1993; Owen et al., 1993;

Hayes et al., 1998) or switching between tasks (Cools et al., 2001b),

this impairment may re¯ect a dopaminergic de®cit (Hayes et al.,

1998; Cools et al., 2001a,b), a dif®culty to instantiate a new

attentional set (Owen et al., 1993), or may be related to the disrupted

forms of motor output (Robertson & Flowers, 1990). The dopami-

nergic de®cit interpretation seems especially pertinent because

several studies reported improved performance in Parkinson's disease

patients after dopaminergic medication (Hayes et al., 1998; Cools

et al., 2001a).

When compared to baseline, all task-switching conditions averaged

together not only activated the cerebellum and the striatum but also

recruited a bilateral prefronto-parietal network (Table 1, Fig. 9a).

This network is typically reported in functional neuroimaging

experiments using attention tasks, working memory tasks and dual-

tasks (Cohen et al., 1997; Nobre et al., 1997; Adcock et al., 2000;

Bunge et al., 2000), indicating that this network is not speci®c to task

switching.

Task order unpredictability activates the anterior striatum

Increased response times on both the switch and repeat trials of the

unpredictable task order conditions show that unpredictability was

related to the overall structure of the sequence of tasks. This con®rms

a recent behavioural study (Sohn & Carlson, 2000). The anterior

putamen activation found with task order unpredictability (Fig. 5) is

in accordance with the observation that striatal neurons detect

unpredicted reward events, irrespective of the speci®c behavioural

situation in which such events occur (Ravel et al., 2001). A previous

fMRI study has shown that unpredictability of reward delivery in

humans activates the ventral striatum (Berns et al., 2001) but could

not specify whether this activation was speci®c to the unpredictability

of the reward delivery or to the unpredictability of the sequence itself.

Our results suggest that reward delivery is unnecessary for activation

of the anterior striatum and that unpredictability of task order alone is

suf®cient to induce this activation. However, the role of dopamine

cannot be totally excluded in our study, even though no reward was

delivered. Indeed, the discharges of midbrain dopamine neurons have

properties similar to the reward prediction error of temporal

difference reinforcement learning models (Schultz & Dickinson,

2000; Schultz, 2000). Dopamine neuron activity serves as an

effective reinforcement signal for learning sensorimotor associations

in the striatum. It is thus possible that the dopamine signal is required

with unpredictable sequences, especially with combined unpredict-

ability of task order and timing. Con®rming this view, substantia

nigra activation was found when comparing the switching condition

with unpredictable task order and random timing to the baseline

(Fig. 8).

When considering other brain regions activated with task order

unpredictability, we found a large network including the right inferior

frontal gyrus (BA 45), the left middle frontal gyrus (BA 9), the

intraparietal cortex bilaterally and the inferior temporal cortex (BA

20 and BA 37) (Table 1, and see Dreher et al., 2002). This con®rms

that the basal ganglia is not the only brain region coding the error

signals (Schultz & Dickinson, 2000; Schultz, 2000). Our results are

also consistent with recent studies showing that decision-making in

the presence of uncertainty involves a fronto-parietal network (Paulus

et al., 2001; Huettel et al., 2002) and that the dorsolateral prefrontal

cortex is associated with the adjustment of inferential learning on the

basis of unpredictability during a causal associative learning task

(Fletcher et al., 2001). At the outset, when all associations were

unpredictable, DLPFC activation was maximal. This response



FIG. 9. (a) Pattern of brain activation in the different task-switching conditions relative to baseline (Z > 3.09, i.e. P < 0.001, uncorrected). (b) Results of thefunctional connectivity analysis. (Top) Matrix of interregional correlations in the different task-switching conditions. The correlation coef®cients R2 arerepresented as colour gradations. (Bottom) To summarize the results in a lateral view of the brain, data from the left and right hemispheres were collapsedtogether. The connection between two brain regions indicate correlation coef®cients R2 > 0.5 when collapsing together the two hemispheres across the brainregions found in all task-switching conditions relative to baseline. (c) Hypothetical processes performed by the components of the bilateral prefronto-parietalnetwork. The cerebellum is assumed to compute timing adjustment and the basal ganglia to react to task order unpredictability.

FIG. 8. (Left) Substantia nigra activation found in the task-switchingcondition with unpredictable task order and random timing relative tobaseline superimposed on a normalized axial MRI slice averaged acrosssubjects. (Right) Percentage of signal change relative to baseline for eachtask-switching condition. Error bars indicate SEM.



attenuated with learning but, subsequently, activation in the DLPFC

was evoked by surprise violations of the learned association.

Cerebellar hemisphere activation with timing adjustment

Our behavioural analysis of the random-timing task-switching

conditions suggests that a timing operation was elicited by the ISI

manipulation. Indeed, there was an important trend toward a

reduction of RTs with time lag, suggesting that subjects could use

the probabilistic information conveyed by the passage of time to

predict the likelihood of stimulus presentation (i.e. the more time the

subjects had, the more probable the stimulus occurrence). A similar

result has recently been reported (Meiran et al., 2000). Therefore, the

increased cerebellar activation found in random relative to ®xed

timing may be attributed to the fact that the cerebellum tries to predict

the timing of stimulus presentation, yet fails to do it for short ISIs

(resulting in a prediction error because the actual outcome at short ISI

differs from the predicted timing outcome).

However, we can't rule out the possibility that, even if the trials

were blocked by duration of ISI, subjects would take longer to

perform the tasks at short ISIs than at long ones. Thus, it is possible

that the longer RT for short ISIs may be caused by a refractory period

in task performance rather than a lack of time necessary to make a

prediction about stimulus onset (Meiran et al., 2000). At the same

time, subjects may anticipate (predict) when random events will

happen and look for a pattern in random timing onset.

There is a large body of evidence that is consistent with the

hypothesis that the posterior part of the cerebellum is related to

timing adjustment (Ivry et al., 1988; Ivry, 1997). Bilateral lesions of

the lateral part of the posterior lobe of the cerebellum induce de®cits

in monitoring and reproducing timing (Ivry et al., 1988; Nichelli

et al., 1996; Ivry, 1997). Studies of eyeblink conditioning also

showed that the posterior cerebellum plays a critical role in precise

timing adjustment (Yeo & Hardiman, 1992; Gruart & Yeo, 1995).

Similarly, the lateral cerebellar cortex and cerebellar vermis

contribute to a supramodal (auditory and visual) production of a

timed motor response, particularly when it is novel or complex

(Penhune et al., 1998). Although these studies support the cerebellar

timing hypothesis, they provide little insight into what particular

regions within the cerebellar hemispheres are most critical for timing

processes. Our results show that the right cerebellar hemisphere

(lobules VI/VII/Crus I) may be particularly suited to perform this

function. Moreover, the posterior cerebellar activation which we

found is consistent with a division of the cerebellum made according

to a rostro-caudal axis. The posterior lobe, especially HVI±HVIIa,

has been linked to such higher order functions as attention and

working memory (Allen et al., 1997; Desmond et al., 1997), while the

anterior lobe represents movement execution. The absence of anterior

cerebellar hemisphere activation in our study is explained by the fact

that the same motor responses were needed in all tasks, and were

subtracted in the contrast.

The timing adjustment hypothesis may be considered as a

particular case of a more general error correction process.

However, our results show the distinct nature of the signal processed

by the cerebellum and the basal ganglia: the cerebellum adjusts to

timing information while the striatum adjusts to task order predict-

ability (Fig. 9c). This distinction is relative for the striatum because

the head of the caudate nucleus and bilateral putamen showed an

interaction between timing and task order (Fig. 7).

It should also be noted that the cerebellum does not work in

isolation, despite the fact that no other brain region than the

cerebellum was activated with random relative to ®xed timing

(Fig. 9b). Indeed, our functional connectivity analysis showed that

when timing was random (and task order was predictable or

unpredictable), the cerebellar activation positively correlated with

the DLPFC, the IPS and the caudate activation. Thus, the cerebellum

is likely to share timing information with these brain regions, which is

consistent with their known anatomical connectivity (Middleton &

Strick, 2000; Clower et al., 2001; Middleton & Strick, 2001).

An alternative view of the discrimination of temporal intervals, the

basis of prediction in the random timing condition, is to consider it a

®ne-scale discrimination of sensory information (Gao et al., 1996). If

so, then cerebellar activity could, in part, re¯ect the sensory

processing necessary to predict time onset. Finally, it may be noted

that although the cerebellar neural circuitry involved in storage of

memories for learned motor responses (e.g. eyelid conditioning) is

now well known (Medina et al., 2002), the exact mechanisms

underlying the cerebellar timing function remain unclear. A recent

study indicated that the conditioned response expression and timing

are dissociable and involve different inhibitory inputs (Bao et al.,

2002).

Comparing the neural basis of sequences of tasks andsequences of movements

Several studies support distinct roles for the basal ganglia and the

cerebellum in learning motor sequences (Jueptner & Weiller, 1998;

Doya, 2000). Doya and collaborators suggested that the cerebellum is

involved in supervised learning and the basal ganglia in reinforce-

ment learning. Jueptner proposed a double dissociation between

selection of movements, which requires the basal ganglia but not the

cerebellum, and sensory information processing, which involves the

cerebellum but not the basal ganglia. Unlike these studies comparing

the roles of the basal ganglia and the cerebellum in motor learning,

our study is the ®rst to directly compare the roles of the basal ganglia

and of the cerebellum in processing sequences of cognitive tasks. The

main difference between motor sequences and cognitive sequences is

that, in motor sequences, one act leads to the next in a chain-like

fashion. In contrast, during sequences of tasks, decisions (e.g.

between two motor responses) need to be taken following rules of the

type: `if the letter is red, is it a vowel'? Unlike previous studies

comparing the roles of the basal ganglia and the cerebellum in motor

learning, our results suggest a distinction in the pattern of brain

activation between unpredictability of task order and timing irregu-

larity.

The question of whether the factors of order and timing involve the

same neural basis for sequences of movements and for sequences of

tasks is still unanswered. A 2 3 2 factorial design varying task order

and timing predictability has previously been used for sequences of

simple ®nger movements cued by the colour of two stimuli (Sakai

et al., 2000). It is, however, dif®cult to directly compare our results

with those of this study because no analysis of the main effects of

stimulus order and timing irregularity was performed. This study

reported bilateral posterior cerebellar hemisphere activation when the

timing was random and no cerebellar activation with unpredictability

of the order of the movement (`response selection'). In addition, no

basal ganglia activation was found with response selection uncer-

tainty (Sakai et al., 2000). However, the role of the basal ganglia in

unpredictable sequences of movements requires further investigation.

Indeed, at the beginning of learning a motor sequence the anterior

striatum is activated (Grafton et al., 1992; Petersen et al., 1998;

Hikosaka et al., 1999). This activation decreased when motor

sequence become more automatic while activation of the posterior

striatum increased. Similarly in our study, when task order cannot be

learned (unpredictable order), the anterior striatum was activated

while a decrease of activation was found when task order was



overlearned predictable task order. This suggests that both motor and

task sequences may require the anterior striatum when they are

unpredictable. Further studies directly comparing motor and cogni-

tive sequences are needed to solve this issue (Koechlin et al., 2002)

To conclude, our study provides new evidence that the cerebellum

and striatum are involved not only in the control of movements but

also in the sequencing of cognitive tasks. The cerebellum is primarily

activated by timing irregularity while the basal ganglia show a more

complex pattern of activation, the anterior striatum being activated by

task order unpredictability while the head of the caudate nucleus and

the posterior putamen responding to an interaction between timing

and task order. Furthermore, task order unpredictability alone,

independently of reward delivery, is suf®cient to induce striatal

activation.

Acknowledgement

This research was supported by a postdoctoral fellowship from the FYSSENfoundation to J.-C.D.

Abbreviations

DLPFC, dorsolateral prefrontal cortex; fMRI, functional magnetic resonanceimaging; IPS, intraparietal sulcus; ISI, interstimulus interval; pre-SMA, pre-supplementary motor area; RT, reaction time.

References

Adcock, R.A., Constable, R.T., Gore, J.C. & Goldman-Rakic, P.S. (2000)Functional neuroanatomy of executive processes involved in dual-taskperformance. Proc. Natl Acad. Sci. USA, 97, 3567±3572.

Akshoomoff, N.A. & Courchesne, E. (1992) A new role for the cerebellum incognitive operations. Behav. Neurosci., 106, 731±738.

Albus, J.S. (1971) The theory of cerebellar function. Mathemat. Biosciences,10, 25±61.

Allen, G., Buxton, R.B., Wong, E.C. & Courchesne, E. (1997) Attentionalactivation of the cerebellum independent of motor involvement. Science,275, 1940±1943.

Bao, S., Chen, L., Kim, J.J. & Thompson, R.F. (2002) Cerebellar corticalinhibition and classical eyeblink conditioning. Proc. Natl Acad. Sci. USA,99, 1592±1597.

Berns, G.S., McClure, S.M., Pagnoni, G. & Montague, P.R. (2001)Predictability modulates human brain response to reward. J. Neurosci.,21, 2793±2798.

Bischoff-Grethe, A., Ivry, R.B. & Grafton, S.T. (2002) Cerebellar involvementin response reassignment rather than attention. J. Neurosci., 22, 546±553.

Blakemore, S.J., Frith, C.D. & Wolpert, D.M. (2001) The cerebellum isinvolved in predicting the sensory consequences of action. Neuroreport, 12,1879±1884.

Bracke-Tolkmitt, R., Linden, A.G.M., Canavan, B., Rockstroh, E., Scholz, K.,Wessel, H.-C. & Diener, H.C. (1989) The cerebellum contributes to mentalskills. Behav. Neurosci., 103, 442±446.

Bunge, S.A., Klingberg, T., Jacobsen, R.B. & Gabrieli, J.D. (2000) A resourcemodel of the neural basis of executive working memory. Proc. Natl Acad.Sci. USA, 97, 3573±3578.

Canavan, A.G., Sprengelmeyer, R., Diener, H.C. & Homberg, V. (1994)Conditional associative learning is impaired in cerebellar disease in humans.Behav. Neurosci., 108, 475±485.

Clark, V.P., Fannon, S., Lai, S., Benson, R. & Bauer, L. (2000) Responses torare visual target and distractor stimuli using event-related fMRI. J.Neurophysiol., 83, 3133±3139.

Clower, D.M., West, R.A., Lynch, J.C. & Strick, P.L. (2001) The inferiorparietal lobule is the target of output from the superior colliculus,hippocampus, and cerebellum. J. Neurosci., 21, 6283±6291.

Cohen, J.D., Perlstein, W.M., Braver, T.S., Nystrom, L.E., Noll, D.C., Jonides,J. & Smith, E.E. (1997) Temporal dynamics of brain activation during aworking memory task [see comments]. Nature, 386, 604±608.

Cools, R., Barker, R.A., Sahakian, B.J. & Robbins, T.W. (2001a) Enhanced or

impaired cognitive function in Parkinson's disease as a function ofdopaminergic medication and task demands. Cereb. Cortex, 11, 1136±1143.

Cools, R., Barker, R.A., Sahakian, B.J. & Robbins, T.W. (2001b) Mechanismsof cognitive set ¯exibility in Parkinson's disease. Brain, 124, 2503±2512.

Courchesne, E. & Allen, G. (1997) Prediction and preparation, fundamentalfunctions of the cerebellum. Learn. Mem., 4, 1±35.

Courchesne, E., Townsend, J., Akshoomoff, N.A., Saitoh, O., Yeung-Courchesne, R., Lincoln, A.J., James, H.E., Haas, R.H., Schreibman, L. &Lau, L. (1994) Impairment in shifting attention in autistic and cerebellarpatients. Behav. Neurosci., 108, 848±865.

Desmond, J.E., Gabrieli, J.D., Wagner, A.D., Ginier, B.L. & Glover, G.H.(1997) Lobular patterns of cerebellar activation in verbal working-memoryand ®nger-tapping tasks as revealed by functional MRI. J. Neurosci., 17,9675±9685.

Dimitrov, M., Grafman, J., Kosseff, P., Wachs, J., Alway, D., Higgins, J.,Litvan, I., Lou, J.S. & Hallett, M. (1996) Preserved cognitive processes incerebellar degeneration. Behav. Brain Res., 79, 131±135.

Dove, A., Pollmann, S., Schubert, T., Wiggins, C.J. & von Cramon, D.Y.(2000) Prefrontal cortex activation in task switching: an event-related fMRIstudy. Brain Res. Cogn. Brain Res., 9, 103±109.

Doya, K. (2000) Complementary roles of basal ganglia and cerebellum inlearning and motor control. Curr. Opin. Neurobiol., 10, 732±739.

Dreher, J.C., Koechlin, E., Ali, S.O. & Grafman, J. (2002) The roles of timingand task order during task switching. Neuroimage, 17, 95±109.

Eslinger, P.J. & Grattan, L.M. (1993) Frontal lobe and frontal-striatalsubstrates for different forms of human cognitive ¯exibility.Neuropsychologia, 31, 17±28.

Fiez, J.A., Petersen, S.E., Cheney, M.K. & Raichle, M.E. (1992) Impairednon-motor learning and error detection associated with cerebellar damage.A single case study. Brain, 115, 155±178.

Fletcher, P.C., Anderson, J.M., Shanks, D.R., Honey, R., Carpenter, T.A.,Donovan, T., Papadakis, N. & Bullmore, E.T. (2001) Responses of humanfrontal cortex to surprising events are predicted by formal associativelearning theory. Nature Neurosci., 4, 1043±1048.

Friston, K.J., Frith, C.D., Liddle, P.F. & Frackowiak, R.S. (1991) Comparingfunctional (PET) images: the assessment of signi®cant change. J. Cereb.Blood Flow Metab., 11, 690±699.

Gao, J.H., Parsons, L.M., Bower, J.M., Xiong, J., Li, J. & Fox, P.T. (1996)Cerebellum implicated in sensory acquisition and discrimination rather thanmotor control. Science, 272, 545±547.

Gotham, A.M., Brown, R.G. & Marsden, C.D. (1988) `Frontal' cognitivefunction in patients with Parkinson's disease `on' and `off' levodopa. Brain,111, 299±321.

Grafton, S.T., Mazziotta, J.C., Woods, R.P. & Phelps, M.E. (1992) Humanfunctional anatomy of visually guided ®nger movements. Brain, 115, 565±587.

Gruart, A. & Yeo, C.H. (1995) Cerebellar cortex and eyeblink conditioning:bilateral regulation of conditioned responses. Exp. Brain Res., 104, 431±448.

Harrington, D.L., Haaland, K.Y. & Knight, R.T. (1998) Cortical networksunderlying mechanisms of time perception. J. Neurosci., 18, 1085±1095.

Hayes, A.E., Davidson, M.C., Keele, S.W. & Rafal, R.D. (1998) Toward afunctional analysis of the basal ganglia. J. Cogn. Neurosci., 10, 178±198.

Helmuth, L.L., Ivry, R.B. & Shimizu, N. (1997) Preserved performance bycerebellar patients on tests of word generation, discrimination learning, andattention. Learn. Mem., 3, 456±474.

Hikosaka, O., Nakahara, H., Rand, M.K., Sakai, K., Lu, X., Nakamura, K.,Miyachi, S. & Doya, K. (1999) Parallel neural networks for learningsequential procedures. Trends Neurosci., 22, 464±471.

Huettel, S.A., Mack, P.B. & McCarthy, G. (2002) Perceiving patterns inrandom series: dynamic processing of sequence in prefrontal cortex. NatureNeurosci., 5, 485±490.

Imamizu, H., Miyauchi, S., Tamada, T., Sasaki, Y., Takino, R., Putz, B.,Yoshioka, T. & Kawato, M. (2000) Human cerebellar activity re¯ecting anacquired internal model of a new tool. Nature, 403, 192±195.

Ivry, R. (1996) The representation of temporal information in perception andmotor control. Curr. Opin. Neurobiol., 6, 851±857.

Ivry, R. (1997) Cerebellar timing systems. Int. Rev. Neurobiol., 41, 555±573.Ivry, R. & Fiez, J.A. (2001) Cerebellar contributions to cognition and imagery.

In Gazzaniga, M.S. (ed.), The New Cognitive Neuroscience. MIT Press,Cambridge, MA, pp 999±1011.

Ivry, R.B., Keele, S.W. & Diener, H.C. (1988) Dissociation of the lateral andmedial cerebellum in movement timing and movement execution. Exp.Brain Res., 73, 167±180.

Jueptner, M. & Weiller, C. (1998) A review of differences between basal



ganglia and cerebellar control of movements as revealed by functionalimaging studies. Brain, 121, 1437±1449.

Kawato, M. & Gomi, H. (1992) A computational model of four regions of thecerebellum based on feedback-error learning. Biol. Cybern, 68, 95±103.

Kimberg, D.Y., Aguirre, G.K. & D'Esposito, M. (2000) Modulation of task-related neural activity in task-switching: an fMRI study (1). Brain Res.Cogn. Brain Res., 10, 189±196.

Koechlin, E., Danek, A., Burnod, Y. & Grafman, J. (2002) Medial prefrontaland subcortical mechanisms underlying the acquisition of motor andcognitive action sequences in humans. Neuron, 35, 371±381.

Le, T.H., Pardo, J.V. & Hu, X. (1998) 4 T-fMRI study of nonspatial shifting ofselective attention: cerebellar and parietal contributions. J. Neurophysiol.,79, 1535±1548.

Leiner, H.C., Leiner, A.L. & Dow, R.S. (1991) The human cerebro-cerebellarsystem: its computing, cognitive, and language skills. Behav. Brain Res., 44,113±128.

Linden, D.E., Prvulovic, D., Formisano, E., Vollinger, M., Zanella, F.E.,Goebel, R. & Dierks, T. (1999) The functional neuroanatomy of targetdetection: an fMRI study of visual and auditory oddball tasks. Cereb.Cortex, 9, 815±823.

Marr, D. (1969) A theory of cerebellar cortex. J. Physiol. (Lond.), 202, 437±470.

Medina, J.F., Christopher Repa, J., Mauk, M.D. & LeDoux, J.E. (2002)Parallels between cerebellum- and amygdala-dependent conditioning.Nature Rev. Neurosci., 3, 122±131.

Meiran, N., Chorev, Z. & Sapir, A. (2000) Component processes in taskswitching. Cognit. Psychol., 41, 211±253.

Miall, R.C., Reckess, G.Z. & Imamizu, H. (2001) The cerebellum coordinateseye and hand tracking movements. Nature Neurosci., 4, 638±644.

Middleton, F.A. & Strick, P.L. (2000) Basal ganglia and cerebellar loops:motor and cognitive circuits. Brain Res. Brain Res. Rev., 31, 236±250.

Middleton, F.A. & Strick, P.L. (2001) Cerebellar projections to the prefrontalcortex of the primate. J. Neurosci., 21, 700±712.

Nichelli, P., Alway, D. & Grafman, J. (1996) Perceptual timing in cerebellardegeneration. Neuropsychologia, 34, 863±871.

Nobre, A.C., Sebestyen, G.N., Gitelman, D.R., Mesulam, M.M., Frackowiak,R.S. & Frith, C.D. (1997) Functional localization of the system forvisuospatial attention using positron emission tomography. Brain, 120, 515±533.

Owen, A.M., Roberts, A.C., Hodges, J.R., Summers, B.A., Polkey, C.E. &Robbins, T.W. (1993) Contrasting mechanisms of impaired attentional set-shifting in patients with frontal lobe damage or Parkinson's disease. Brain,116, 1159±1175.

Pastor, M.A., Artieda, J., Jahanshahi, M. & Obeso, J.A. (1992) Timeestimation and reproduction is abnormal in Parkinson's disease. Brain, 115,211±225.

Paulus, M.P., Hozack, N., Zauscher, B., McDowell, J.E., Frank, L., Brown,G.G. & Braff, D.L. (2001) Prefrontal, parietal, and temporal cortexnetworks underlie decision-making in the presence of uncertainty.Neuroimage, 13, 91±100.

Penhune, V.B., Zattore, R.J. & Evans, A.C. (1998) Cerebellar contributions to

motor timing: a PET study of auditory and visual rhythm reproduction. J.Cogn. Neurosci., 10, 752±765.

Petersen, S.E., van Mier, H., Fiez, J.A. & Raichle, M.E. (1998) The effects ofpractice on the functional anatomy of task performance. Proc. Natl Acad.Sci. USA, 95, 853±860.

Ravel, S., Sardo, P., Legallet, E. & Apicella, P. (2001) Rewardunpredictability inside and outside of a task context as a determinant ofthe responses of tonically active neurons in the monkey striatum. J.Neurosci., 21, 5730±5739.

Ravizza, S.M. & Ivry, R.B. (2001) Comparison of the basal ganglia andcerebellum in shifting attention. J. Cogn. Neurosci., 13, 285±297.

Robertson, C. & Flowers, K.A. (1990) Motor set in Parkinson's disease. J.Neurol. Neurosurg. Psychiatry, 53, 583±592.

Rogers, R.D., Sahakian, B.J., Hodges, J.R., Polkey, C.E., Kennard, C. &Robbins, T.W. (1998) Dissociating executive mechanisms of task controlfollowing frontal lobe damage and Parkinson's disease. Brain, 121, 815±842.

Sakai, K., Hikosaka, O., Takino, R., Miyauchi, S., Nielsen, M. & Tamada, T.(2000) What and when: parallel and convergent processing in motorControl. J. Neurosci., 20, 2691±2700.

Schmahmann, J.D., Doyon, J., McDonald, D., Holmes, C., Lavoie, K.,Hurwitz, A.S., Kabani, N., Toga, A., Evans, A.C. & Petrides, M. (1999)Three-dimensional MRI atlas of the human cerebellum in proportionalstereotaxic space. Neuroimage, 10, 233±260.

Schultz, W. (2000) Multiple reward signals in the brain. Nature Rev.Neurosci., 1, 199±207.

Schultz, W. & Dickinson, A. (2000) Neuronal coding of prediction errors.Annu. Rev. Neurosci., 23, 473±500.

Sohn, M.H. & Carlson, R.A. (2000) Effects of repetition and foreknowledge intask-set recon®guration. J. Exp. Psychol. Learn. Mem. Cogn., 26, 1445±1460.

Sohn, M.H., Ursu, S., Anderson, J.R., Stenger, V.A. & Carter, C.S. (2000) Therole of prefrontal cortex and posterior parietal cortex in task switching.Proc. Natl Acad. Sci. USA, 97, 13448±13453.

Stevens, A.A., Skudlarski, P., Gatenby, J.C. & Gore, J.C. (2000) Event-relatedfMRI of auditory and visual oddball tasks. Magn. Reson. Imaging, 18, 495±502.

Tamada, T., Miyauchi, S., Imamizu, H., Yoshioka, T. & Kawato, M. (1999)Cerebro-cerebellar functional connectivity revealed by the laterality indexin tool-use learning. Neuroreport, 10, 325±331.

Taylor, A.E., Saint-Cyr, J.A. & Lang, A.E. (1986) Frontal lobe dysfunction inParkinson's disease. The cortical focus of neostriatal out¯ow. Brain, 109,845±883.

Thach, W.T. (1998) A role for the cerebellum in learning movementcoordination. Neurobiol. Learn. Mem., 70, 177±188.

Yamaguchi, S., Tsuchiya, H. & Kobayashi, S. (1998) Visuospatial attentionshift and motor responses in cerebellar disorders. J. Cogn. Neurosci., 10,95±107.

Yeo, C.H. & Hardiman, M.J. (1992) Cerebellar cortex and eyeblinkconditioning: a reexamination. Exp Brain Res., 88, 623±638.



The roles of the cerebellum and basal ganglia in timing and error prediction

Documents

Transcript of The roles of the cerebellum and basal ganglia in timing and error prediction